Upload g2_no_memory_s1.json with huggingface_hub
Browse files- g2_no_memory_s1.json +33 -11
g2_no_memory_s1.json
CHANGED
|
@@ -275,29 +275,51 @@
|
|
| 275 |
"lr": 0.0003,
|
| 276 |
"eta": 0.0,
|
| 277 |
"s_per_step": 0.0502
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 278 |
}
|
| 279 |
],
|
| 280 |
"final": {
|
| 281 |
"law_rev": "candm-v1",
|
| 282 |
-
"value_acc_d32": 0.
|
| 283 |
"recall_exact_d32": 0.0,
|
| 284 |
-
"key_acc_d32": 0.
|
| 285 |
"value_acc_d128": 0.0039,
|
| 286 |
"recall_exact_d128": 0.0,
|
| 287 |
-
"key_acc_d128": 0.
|
| 288 |
-
"value_acc_d512": 0.
|
| 289 |
"recall_exact_d512": 0.0,
|
| 290 |
-
"key_acc_d512": 0.
|
| 291 |
-
"value_acc_d1024": 0.
|
| 292 |
"recall_exact_d1024": 0.0,
|
| 293 |
-
"key_acc_d1024": 0.
|
| 294 |
-
"step":
|
| 295 |
-
"train_bpb": 7.
|
| 296 |
-
"grad_norm":
|
| 297 |
"pairs": 4,
|
| 298 |
"lr": 0.0003,
|
| 299 |
"eta": 0.0,
|
| 300 |
-
"s_per_step": 0.
|
| 301 |
}
|
| 302 |
}
|
| 303 |
],
|
|
|
|
| 275 |
"lr": 0.0003,
|
| 276 |
"eta": 0.0,
|
| 277 |
"s_per_step": 0.0502
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"law_rev": "candm-v1",
|
| 281 |
+
"value_acc_d32": 0.0039,
|
| 282 |
+
"recall_exact_d32": 0.0,
|
| 283 |
+
"key_acc_d32": 0.0065,
|
| 284 |
+
"value_acc_d128": 0.0039,
|
| 285 |
+
"recall_exact_d128": 0.0,
|
| 286 |
+
"key_acc_d128": 0.0052,
|
| 287 |
+
"value_acc_d512": 0.0065,
|
| 288 |
+
"recall_exact_d512": 0.0,
|
| 289 |
+
"key_acc_d512": 0.0065,
|
| 290 |
+
"value_acc_d1024": 0.0065,
|
| 291 |
+
"recall_exact_d1024": 0.0,
|
| 292 |
+
"key_acc_d1024": 0.0026,
|
| 293 |
+
"step": 5500,
|
| 294 |
+
"train_bpb": 7.2346,
|
| 295 |
+
"grad_norm": 3.335874557,
|
| 296 |
+
"pairs": 4,
|
| 297 |
+
"lr": 0.0003,
|
| 298 |
+
"eta": 0.0,
|
| 299 |
+
"s_per_step": 0.051
|
| 300 |
}
|
| 301 |
],
|
| 302 |
"final": {
|
| 303 |
"law_rev": "candm-v1",
|
| 304 |
+
"value_acc_d32": 0.0039,
|
| 305 |
"recall_exact_d32": 0.0,
|
| 306 |
+
"key_acc_d32": 0.0065,
|
| 307 |
"value_acc_d128": 0.0039,
|
| 308 |
"recall_exact_d128": 0.0,
|
| 309 |
+
"key_acc_d128": 0.0052,
|
| 310 |
+
"value_acc_d512": 0.0065,
|
| 311 |
"recall_exact_d512": 0.0,
|
| 312 |
+
"key_acc_d512": 0.0065,
|
| 313 |
+
"value_acc_d1024": 0.0065,
|
| 314 |
"recall_exact_d1024": 0.0,
|
| 315 |
+
"key_acc_d1024": 0.0026,
|
| 316 |
+
"step": 5500,
|
| 317 |
+
"train_bpb": 7.2346,
|
| 318 |
+
"grad_norm": 3.335874557,
|
| 319 |
"pairs": 4,
|
| 320 |
"lr": 0.0003,
|
| 321 |
"eta": 0.0,
|
| 322 |
+
"s_per_step": 0.051
|
| 323 |
}
|
| 324 |
}
|
| 325 |
],
|